Modeling the onset advantage in musical instrument recognition.

Marc René Schädler,David Hülsmeier,Kai Siedenburg

doi:10.1121/1.5141369

Abstract

Sound onsets provide particularly valuable cues for musical instrument identification by human listeners. It has remained unclear whether this onset advantage is due to enhanced perceptual encoding or the richness of acoustical information during onsets. Here this issue was approached by modeling a recent study on instrument identification from tone excerpts [Siedenburg. (2019). J. Acoust. Soc. Am. 145(2), 1078-1087]. A simple Hidden Markov Model classifier with separable Gabor filterbank features simulated human performance and replicated the onset advantage observed previously for human listeners. These results provide evidence that the onset advantage may be driven by the distinct acoustic qualities of onsets.

Highlights

The identification of musical instruments is a central task in music perception (e.g., Rentfrow and Levitin, 2019)
Accuracies for classifiers trained on the full 250 ms sounds are depicted in Fig. 2, together with experimental results from human listeners
Both the separable Gabor filterbank (SGBFB) and Mel-frequency cepstral coefficients (MFCCs) features still provided a pattern of results that qualitatively resembled that of human listeners

Summary

Introduction

The identification of musical instruments is a central task in music perception (e.g., Rentfrow and Levitin, 2019). A landmark effect concerns sound onsets, which are suspected to provide valuable cues for instrument identification: if presented with sound excerpts, human listeners more identify instrument sounds from onset portions compared to other portions of the sound (Saldanha and Corso, 1964; Schaeffer, 2017)—a behavioral effect that we refer to as onset advantage. This effect does not imply that all instrumental sounds become unidentifiable without onsets, because informative cues may be extracted across the full sound duration and the degree to which this is possible may depend on the specific instrument at hand (Agus et al, 2019). Auditory modeling approaches are in a position to provide valuable insights into this issue

Methods

Results

Conclusion